US Finalizes Voluntary AI Pre-Release Review Framework, Invites OpenAI and Anthropic

Trump administration completes voluntary safety review protocol for frontier AI models, inviting major firms to discuss cybersecurity and national security risk assessments.

Woofun AI reports that the Trump administration has finalized a voluntary pre-release review framework for frontier AI models, inviting representatives from OpenAI, Google, and Anthropic to the White House for discussions. The mechanism enables companies to voluntarily submit models for government safety testing prior to public release, focusing on cybersecurity capabilities and national security risks without establishing mandatory approval. This framework implements an executive order signed on June 2, which mandated rule formulation within 60 days to identify models requiring priority review. The report notes previous interim measures, including export restrictions on Anthropic's Fable model and phased release requirements for OpenAI's GPT-5.6, suggesting this framework aims to institutionalize such interventions.